2016 INFORMS Annual Meeting Program

WC22

INFORMS Nashville – 2016

4 - Optimizing Admission And Discharge Decisions In Icu With Flexible Bed Allocation Xuanjing Li, Tsinghua University, Room 519A,Shunde Building,Tsinghua Univ, Beijing, Beijing, 100084, China, lixj15@mails.tsinghua.edu.cn, Dacheng Liu, Xiaolei Xie, Ye Wang This paper studies the admission and premature discharge policy in the intensive care unit (ICU) at Peking University Third Hospital. Patients are classified into two categories based on their survival benefit and discharge cost. A Markov Decision Process (MDP) model is established to strike balance between those two factors. Structural properties are obtained and a new admission policy is proposed. WC22 107B-MCC Appointment Scheduling Models and Analytics Sponsored: Health Applications Sponsored Session Chair: Nan Liu, Columbia University, 722 W. 168th. St, New York, NY, 10032, United States, nl2320@columbia.edu Co-Chair: Zhankun Sun, Eyes High postdoctoral scholar, University of Calgary, 2500 University Dr. NW, Calgary, AB, T2N 1N4, Canada, zhankun.sun@haskayne.ucalgary.ca 1 - Improving Patient Satisfaction: Customizing Patient Appointment In this paper, we develop a Bayesian logit model which improves no-show prediction accuracy over the widely-used simple logit model. The accuracy gain arises from the individual patient-level coefficients provided by the Bayesian approach. Comparison of model fit on 12-months of appointment data shows the Bayesian model outperforms the simple logit model. In simulation studies, our results show that applying Bayesian model’s prediction to scheduling algorithm can reduce patient’s waiting time and physician’s idle time and increase clinic’s profit. 2 - Managing Appointment-based Services In The Presence Of Walk-in Customers Shan Wang, Shanghai Jiao Tong University, 1954 Huashan Road, Shanghai, 200030, China, wangshan_731@sjtu.edu.cn, Nan Liu, Guohua Wan Walk-in customers are accepted in many service industries and especially in healthcare. Motivated by practice in and data collected from two large community health care networks in New York City, we study how to coordinate scheduled patients in a clinic session in anticipation for random walk-ins. We use a Poisson regression framework to analyze the temporal pattern of walk-in patients based on 3-year data, and propose data-driven optimization models to identify the optimal appointment schedule. Our models can incorporate other practical aspects of appointment scheduling such as patient no-shows, patient preferences and restricted walk-in windows. 3 - Physician Scheduling To Improve Patient Flow Through Emergency Rooms Farzad Zaerpour, PhD Candidate, Haskayne School of Business, University of Calgary, Calgary, AB, T3A 2E1, Canada, farzad.zaerpour@haskayne.ucalgary.ca, Zhankun Sun, Marco Bijvank Emergency department (ED) crowding has become a serious concern worldwide. Hours of waiting is the main consequence of crowding in emergency departments. In this study, we develop a mixed-integer stochastic program for scheduling physicians to improve patient flow through an emergency department. The operational performance of an emergency department is vulnerable to mismatch between demand and supply. Therefore, the proposed model takes into account the stochastic natures of both demand and supply. We use physician productivity to evaluate the performance of each physician in the emergency department. 4 - When Waiting To See A Doctor Is Less Irritating: Understanding Patient Preferences And Choice Behavior In Appointment Scheduling Nan Liu, Columbia University, 722 W. 168th St., Room 476, New York, NY, 10032, United States, nl2320@columbia.edu, Stacey Finkelstein, Margaret Kruk, David Rosenthal This talk examines patient preferences and choice behavior in scheduling medical appointments. We conduct four discrete choice experiments on two distinct populations and identify several operational attributes that affect patient choice. We observe an interesting gender effect with respect to how patients tradeoff Yutian Li, University of Miami, 421 Jenkins Building, 5250 University Drive, Coral Gables, FL, 33124, United States, ytli@umiami.edu, Joseph Johnson, Yu Tang

speed (delay to care) vs. quality (doctor of choice), and demonstrate that risk- attitudes mediate the impact of gender. As many operational strategies aim to improve patient experience by making tradeoffs between speed and quality, we make suggestions for when managers should intervene and how such interventions might look based on the patient mix and current delay level.

WC23

108-MCC Optimization in Radiation Therapy Treatment Planning Sponsored: Health Applications Sponsored Session Chair: Victor Wu, University of Michigan, 1205 Beal Avenue, Ann Arbor, MI, 48109, United States, vwwu@umich.edu 1 - Deriving Imrt Treatment Plans From Dvh Curves Aaron Babier, University of Toronto, Toronto, ON, Canada, Plan quality is often assessed using dose volume histograms (DVHs), which are a high level representation of a dose distribution. Clinical quality DVHs can be accurately predicted, however their corresponding treatment plans are more challenging to determine. We present an inverse optimization model that can produce treatment plans from DVH curves, with minimum treatment complexity. The model is applied to several clinical head and neck treatment plans, and the outcomes are compared to their corresponding clinical plans. 2 - Evaluation Of Multi-source Treatments For Prostate Brachytherapy Optimized Using An Interior Point Constraint Generation Algorithm Dionne Aleman, University of Toronto, Toronto, ON, Canada, aleman@mie.utoronto.ca, Rachel Mok Tsze Chung, William Song A novel approach to treat prostate cancer using multi-source high-dose-rate brachytherapy is investigated. The effectiveness of different combinations of the radionuclides 192Ir, 60Co, and 169Yb is analyzed. We use an inverse planning interior point algorithm to generate treatment plans for every possible combination of the three sources, and then compare treatment quality to the 192Ir plan. Overall, for the same target coverage, double- and triple-source plans provided better organ-at-risk sparing than the 192Ir plan. 3 - Threshold-driven Optimization For Reference-based Auto-planning Troy Long, University of Texas Southwestern, Dallas, TX, United States, troy.long@utsouthwestern.edu, Steve Jiang, Mingli Chen, Weiguo Lu We study the procedure of reference-based auto-planning for treatment plan optimization. We develop a threshold-driven optimization methodology for automatically generating an intensity-modulated radiation therapy treatment plan that is motivated by a reference dose-volume histogram. The commonly used voxel-based quadratic penalty objective functions have three components: an overdose weight, and underdose weight, and some target dose threshold. The proposed methodology directly relates reference information to threshold values, which influence the optimization in an effective, intuitive way. 4 - Optimal Fractionation With Two Modalities Sevnaz Nourollahi, University of Washington Seattle, sevnaz@uw.edu We introduce an optimal fractionation problem with two modalities. This involves finding the number of treatment sessions and the dose per session administered via each modality. The goal is to maximize the biological effect of such bimodal treatment on the tumor while keeping the toxic effects on nearby normal tissue within tolerable limits. We formulate this problem as a nonconvex quadratically constrained quadratic program. We show that the KKT conditions for this problem reduce to solving a quartic equation. We are thus able to provide an analytical solution to the KKT system. We study properties of the resulting solutions via numerical experiments. ababier@mie.utoronto.ca, Justin James Boutilier, Andrea McNiven, Michael Sharpe, Timothy Chan

434